
!pip install shap
!pip install transformers
!pip install nlp
!pip install torch
!pip install scipy

Collecting shap
  Downloading shap-0.40.0-cp37-cp37m-manylinux2010_x86_64.whl (564 kB)
     |████████████████████████████████| 564 kB 5.4 MB/s 
Requirement already satisfied: packaging>20.9 in /usr/local/lib/python3.7/dist-packages (from shap) (21.3)
Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from shap) (1.19.5)
Requirement already satisfied: scikit-learn in /usr/local/lib/python3.7/dist-packages (from shap) (1.0.1)
Requirement already satisfied: pandas in /usr/local/lib/python3.7/dist-packages (from shap) (1.1.5)
Requirement already satisfied: cloudpickle in /usr/local/lib/python3.7/dist-packages (from shap) (1.3.0)
Requirement already satisfied: scipy in /usr/local/lib/python3.7/dist-packages (from shap) (1.4.1)
Requirement already satisfied: tqdm>4.25.0 in /usr/local/lib/python3.7/dist-packages (from shap) (4.62.3)
Collecting slicer==0.0.7
  Downloading slicer-0.0.7-py3-none-any.whl (14 kB)
Requirement already satisfied: numba in /usr/local/lib/python3.7/dist-packages (from shap) (0.51.2)
Requirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging>20.9->shap) (3.0.6)
Requirement already satisfied: setuptools in /usr/local/lib/python3.7/dist-packages (from numba->shap) (57.4.0)
Requirement already satisfied: llvmlite<0.35,>=0.34.0.dev0 in /usr/local/lib/python3.7/dist-packages (from numba->shap) (0.34.0)
Requirement already satisfied: python-dateutil>=2.7.3 in /usr/local/lib/python3.7/dist-packages (from pandas->shap) (2.8.2)
Requirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.7/dist-packages (from pandas->shap) (2018.9)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.7.3->pandas->shap) (1.15.0)
Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.7/dist-packages (from scikit-learn->shap) (1.1.0)
Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn->shap) (3.0.0)
Installing collected packages: slicer, shap
Successfully installed shap-0.40.0 slicer-0.0.7
Collecting transformers
  Downloading transformers-4.13.0-py3-none-any.whl (3.3 MB)
     |████████████████████████████████| 3.3 MB 5.4 MB/s 
Collecting huggingface-hub<1.0,>=0.1.0
  Downloading huggingface_hub-0.2.1-py3-none-any.whl (61 kB)
     |████████████████████████████████| 61 kB 521 kB/s 
Requirement already satisfied: filelock in /usr/local/lib/python3.7/dist-packages (from transformers) (3.4.0)
Requirement already satisfied: importlib-metadata in /usr/local/lib/python3.7/dist-packages (from transformers) (4.8.2)
Requirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.7/dist-packages (from transformers) (4.62.3)
Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.7/dist-packages (from transformers) (21.3)
Collecting pyyaml>=5.1
  Downloading PyYAML-6.0-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl (596 kB)
     |████████████████████████████████| 596 kB 46.4 MB/s 
Requirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from transformers) (2.23.0)
Collecting sacremoses
  Downloading sacremoses-0.0.46-py3-none-any.whl (895 kB)
     |████████████████████████████████| 895 kB 45.8 MB/s 
Collecting tokenizers<0.11,>=0.10.1
  Downloading tokenizers-0.10.3-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl (3.3 MB)
     |████████████████████████████████| 3.3 MB 38.9 MB/s 
Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.7/dist-packages (from transformers) (2019.12.20)
Requirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.7/dist-packages (from transformers) (1.19.5)
Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.7/dist-packages (from huggingface-hub<1.0,>=0.1.0->transformers) (3.10.0.2)
Requirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging>=20.0->transformers) (3.0.6)
Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata->transformers) (3.6.0)
Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (2.10)
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (2021.10.8)
Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (3.0.4)
Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (1.24.3)
Requirement already satisfied: joblib in /usr/local/lib/python3.7/dist-packages (from sacremoses->transformers) (1.1.0)
Requirement already satisfied: click in /usr/local/lib/python3.7/dist-packages (from sacremoses->transformers) (7.1.2)
Requirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from sacremoses->transformers) (1.15.0)
Installing collected packages: pyyaml, tokenizers, sacremoses, huggingface-hub, transformers
  Attempting uninstall: pyyaml
    Found existing installation: PyYAML 3.13
    Uninstalling PyYAML-3.13:
      Successfully uninstalled PyYAML-3.13
Successfully installed huggingface-hub-0.2.1 pyyaml-6.0 sacremoses-0.0.46 tokenizers-0.10.3 transformers-4.13.0
Collecting nlp
  Downloading nlp-0.4.0-py3-none-any.whl (1.7 MB)
     |████████████████████████████████| 1.7 MB 5.5 MB/s 
Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from nlp) (1.19.5)
Requirement already satisfied: pyarrow>=0.16.0 in /usr/local/lib/python3.7/dist-packages (from nlp) (3.0.0)
Requirement already satisfied: dill in /usr/local/lib/python3.7/dist-packages (from nlp) (0.3.4)
Requirement already satisfied: pandas in /usr/local/lib/python3.7/dist-packages (from nlp) (1.1.5)
Requirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.7/dist-packages (from nlp) (4.62.3)
Collecting xxhash
  Downloading xxhash-2.0.2-cp37-cp37m-manylinux2010_x86_64.whl (243 kB)
     |████████████████████████████████| 243 kB 47.1 MB/s 
Requirement already satisfied: requests>=2.19.0 in /usr/local/lib/python3.7/dist-packages (from nlp) (2.23.0)
Requirement already satisfied: filelock in /usr/local/lib/python3.7/dist-packages (from nlp) (3.4.0)
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->nlp) (2021.10.8)
Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->nlp) (3.0.4)
Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->nlp) (2.10)
Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->nlp) (1.24.3)
Requirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.7/dist-packages (from pandas->nlp) (2018.9)
Requirement already satisfied: python-dateutil>=2.7.3 in /usr/local/lib/python3.7/dist-packages (from pandas->nlp) (2.8.2)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.7.3->pandas->nlp) (1.15.0)
Installing collected packages: xxhash, nlp
Successfully installed nlp-0.4.0 xxhash-2.0.2
Requirement already satisfied: torch in /usr/local/lib/python3.7/dist-packages (1.10.0+cu111)
Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from torch) (3.10.0.2)
Requirement already satisfied: scipy in /usr/local/lib/python3.7/dist-packages (1.4.1)
Requirement already satisfied: numpy>=1.13.3 in /usr/local/lib/python3.7/dist-packages (from scipy) (1.19.5)


from google.colab import drive
drive.mount('/content/drive')

Mounted at /content/drive


import shap
import transformers
import nlp
import torch
import numpy as np
import scipy as sp
import pandas as pd
import pickle
import matplotlib.pyplot as plt

# load a BERT sentiment analysis model
tokenizer = transformers.DistilBertTokenizerFast.from_pretrained("distilbert-base-uncased")
model = transformers.DistilBertForSequenceClassification.from_pretrained(
    "distilbert-base-uncased-finetuned-sst-2-english"
).cuda()

# define a prediction function
def f(x):
    tv = torch.tensor([tokenizer.encode(v, padding='max_length', max_length=500, truncation=True) for v in x]).cuda()
    outputs = model(tv)[0].detach().cpu().numpy()
    scores = (np.exp(outputs).T / np.exp(outputs).sum(-1)).T
    val = sp.special.logit(scores[:,1]) # use one vs rest logit units
    return val

# build an explainer using a token masker
explainer = shap.Explainer(f, tokenizer)

# explain the model's predictions on IMDB reviews
imdb_train = nlp.load_dataset("imdb")["train"]

#use only first 10 reviews
#shap_values = explainer(imdb_train[:10], fixed_context=1)

#use first 100 reviews
shap_values = explainer(imdb_train[:300], fixed_context=1)

Downloading and preparing dataset imdb/plain_text (download: 80.23 MiB, generated: 127.06 MiB, post-processed: Unknown sizetotal: 207.28 MiB) to /root/.cache/huggingface/datasets/imdb/plain_text/1.0.0/76cdbd7249ea3548c928bbf304258dab44d09cd3638d9da8d42480d1d1be3743...

Dataset imdb downloaded and prepared to /root/.cache/huggingface/datasets/imdb/plain_text/1.0.0/76cdbd7249ea3548c928bbf304258dab44d09cd3638d9da8d42480d1d1be3743. Subsequent calls will reuse this data.

Partition explainer:  50%|█████     | 1/2 [00:00<?, ?it/s]Token indices sequence length is longer than the specified maximum sequence length for this model (559 > 512). Running this sequence through the model will result in indexing errors

Partition explainer: 3it [00:25,  6.39s/it]

Partition explainer: 4it [00:37,  8.62s/it]

Partition explainer: 5it [00:48,  9.75s/it]

Partition explainer: 6it [01:00, 10.40s/it]

Partition explainer: 7it [01:12, 10.89s/it]

Partition explainer: 8it [01:23, 11.13s/it]

Partition explainer: 9it [01:36, 11.69s/it]


data_flat = [item for sublist in shap_values.feature_names for item in sublist]
shap_values_flat = list(np.concatenate(shap_values.values).ravel())

sum_dict = {}

for key, val in zip(data_flat, shap_values_flat): 
    sum_dict[key] = sum_dict.get(key, 0) + val 

count_dict = dict((x, data_flat.count(x)) for x in set(data_flat))
mean_dict = dict((k, float(sum_dict[k]) / count_dict[k]) for k in sum_dict)


sample_idx = 6
sample_token = data_flat[sample_idx]

print('Sample token:', sample_token)
print('1st appearance value:', shap_values_flat[sample_idx])
print('Sum shapely value:', sum_dict[sample_token])
print('Nr. occurrences:', count_dict[sample_token])
print('Mean value:', mean_dict[sample_token])

Sample token: a
1st appearance value: 0.2311327963875186
Sum shapely value: 15.2134793382264
Nr. occurrences: 1927
Mean value: 0.007894903652426778


asc_mean_dict = {k: v for k, v in sorted(mean_dict.items(), key=lambda item: item[1])}
desc_mean_dict = {k: v for k, v in sorted(mean_dict.items(), key=lambda item: item[1], reverse = True)}

print('Highest negative impact:\n', list(asc_mean_dict.items())[:10])
print('Highest positive impact:\n', list(desc_mean_dict.items())[:10])

Highest negative impact:
 [('dumped', -4.709905561229299), ('devoid', -3.2381933901978313), ('ignore', -3.190740628173125), ('dump', -3.105963467239014), ('wasted', -2.401778899385703), ('ignorance', -2.324884117096375), ('flaw', -1.81759786424674), ('Stink', -1.4580655319755758), ('unnecessary', -1.4203426167222393), ('loops', -1.3916533375395566)]
Highest positive impact:
 [('expands', 3.683164169664343), ('benefits', 2.950578252991138), ('improves', 2.5204138905289417), ('relevance', 2.2106137249586952), ('charm', 1.752030800039461), ('effectively', 1.6029766657180793), ('precursor', 1.5250277421864262), ('opens', 1.5119095327317338), ('consistent', 1.4608564965416004), ('tilt', 1.3162544832833507)]


#only run when dicts need to be updated
'''
sum_dict_file = open("/content/drive/MyDrive/XAI/sum_dict.pkl", "wb")
pickle.dump(sum_dict, sum_dict_file)
sum_dict_file.close()

count_dict_file = open("/content/drive/MyDrive/XAI/count_dict.pkl", "wb")
pickle.dump(count_dict, count_dict_file)
count_dict_file.close()

mean_dict_file = open("/content/drive/MyDrive/XAI/mean_dict.pkl", "wb")
pickle.dump(mean_dict, mean_dict_file)
mean_dict_file.close()

asc_mean_dict_file = open("/content/drive/MyDrive/XAI/asc_mean_dict.pkl", "wb")
pickle.dump(asc_mean_dict, asc_mean_dict_file)
asc_mean_dict_file.close()

desc_mean_dict_file = open("/content/drive/MyDrive/XAI/desc_mean_dict.pkl", "wb")
pickle.dump(desc_mean_dict, desc_mean_dict_file)
desc_mean_dict_file.close()'''

'\nsum_dict_file = open("/content/drive/MyDrive/XAI/sum_dict.pkl", "wb")\npickle.dump(sum_dict, sum_dict_file)\nsum_dict_file.close()\n\ncount_dict_file = open("/content/drive/MyDrive/XAI/count_dict.pkl", "wb")\npickle.dump(count_dict, count_dict_file)\ncount_dict_file.close()\n\nmean_dict_file = open("/content/drive/MyDrive/XAI/mean_dict.pkl", "wb")\npickle.dump(mean_dict, mean_dict_file)\nmean_dict_file.close()\n\nasc_mean_dict_file = open("/content/drive/MyDrive/XAI/asc_mean_dict.pkl", "wb")\npickle.dump(asc_mean_dict, asc_mean_dict_file)\nasc_mean_dict_file.close()\n\ndesc_mean_dict_file = open("/content/drive/MyDrive/XAI/desc_mean_dict.pkl", "wb")\npickle.dump(desc_mean_dict, desc_mean_dict_file)\ndesc_mean_dict_file.close()'


sum_dict_file = open("/content/drive/MyDrive/XAI/sum_dict.pkl", "rb")
sum_dict = pickle.load(sum_dict_file)

count_dict_file = open("/content/drive/MyDrive/XAI/count_dict.pkl", "rb")
count_dict = pickle.load(count_dict_file)

mean_dict_file = open("/content/drive/MyDrive/XAI/mean_dict.pkl", "rb")
mean_dict = pickle.load(mean_dict_file)

asc_mean_dict_file = open("/content/drive/MyDrive/XAI/asc_mean_dict.pkl", "rb")
asc_mean_dict = pickle.load(asc_mean_dict_file)

desc_mean_dict_file = open("/content/drive/MyDrive/XAI/desc_mean_dict.pkl", "rb")
desc_mean_dict = pickle.load(desc_mean_dict_file)


#loaded dicts
print('Highest negative impact:\n', list(asc_mean_dict.items())[:10])
print('Highest positive impact:\n', list(desc_mean_dict.items())[:10])

Highest negative impact:
 [('dumped', -4.709905561229299), ('devoid', -3.2381933901978313), ('ignore', -3.190740628173125), ('dump', -3.105963467239014), ('wasted', -2.401778899385703), ('ignorance', -2.324884117096375), ('flaw', -1.81759786424674), ('Stink', -1.4580655319755758), ('unnecessary', -1.4203426167222393), ('loops', -1.3916533375395566)]
Highest positive impact:
 [('expands', 3.683164169664343), ('benefits', 2.950578252991138), ('improves', 2.5204138905289417), ('relevance', 2.2106137249586952), ('charm', 1.752030800039461), ('effectively', 1.6029766657180793), ('precursor', 1.5250277421864262), ('opens', 1.5119095327317338), ('consistent', 1.4608564965416004), ('tilt', 1.3162544832833507)]


top_asc_vals = []
top_asc_names = []

top_desc_vals = []
top_desc_names = []

top_x = 10

for x in list(asc_mean_dict.items())[:top_x]:
    top_asc_names.append(x[0])
    top_asc_vals.append(x[1])

for x in list(desc_mean_dict.items())[:top_x]:
    top_desc_names.append(x[0])
    top_desc_vals.append(x[1])


plot_asc_vals = sorted([abs(ele) for ele in top_asc_vals])
plot_asc_names = list(reversed(top_asc_names))
plot_desc_vals = sorted([abs(ele) for ele in top_desc_vals])
plot_desc_names = list(reversed(top_desc_names))


cmap_blue = plt.get_cmap("Blues")
rescale = lambda y: 0.3 + (y - np.min(y)) / (np.max(y) - np.min(y))

plt.figure(figsize = (10, 5))
plt.barh(plot_asc_names, plot_asc_vals, color = cmap_blue(rescale(plot_asc_vals)))
plt.ylabel("Bert Token") 
plt.xlabel("Negative Impact on Sentiment (Shapely Value)")
plt.title("Top 10 Bert Tokens sorted by Negative Impact on Sentiment")

plt.show()


cmap_red = plt.get_cmap("Reds")
rescale = lambda y: 0.3 + (y - np.min(y)) / (np.max(y) - np.min(y))

plt.figure(figsize = (10, 5))
plt.barh(plot_desc_names, plot_desc_vals, color = cmap_red(rescale(plot_asc_vals)))
plt.ylabel("Bert Token") 
plt.xlabel("Positive Impact on Sentiment (Shapely Value)")
plt.title("Top 10 Bert Tokens sorted by Positive Impact on Sentiment")

plt.show()


# plot the first sentence's explanation
shap.plots.text(shap_values[4])


shap.plots.text(shap_values)